Dehazing Method of Traffic Image Based on Contour Wave and Markov Random Field

Abstract

The invention discloses a traffic image defogging method based on non-subsampling contour wave and Markov random field. First, the foggy image is converted from RGB color space to HSV space, and non-subsampled contourlet transformation is performed on it; secondly, the high-frequency contourlet coefficients are nonlinearly transformed to enhance the detail information, and the foggy image with edge enhancement is obtained. ;Thirdly, use the low-frequency contourlet coefficients to estimate the atmospheric light value, and establish a Markov random field model on the low-frequency contourlet subband, and use the graph-cut-based α-extension method and bilateral filtering method to calculate the transmittance; finally, use the atmospheric light Value, transmittance and atmospheric scattering model, and restore the fogged image after edge enhancement to obtain the HSV image after defogging, and then convert it to RGB color space as the final output.

Description

technical field [0001] The invention relates to the field of intelligent traffic video processing, in particular to a traffic image defogging method based on non-subsampling contourlet and Markov random field, which has strong edge preservation ability, good robustness, high contrast and high definition. Background technique: [0002] Although the intelligent traffic management system can obtain vehicle information (such as extracting vehicle model, license plate number and body color, etc.) with the help of a large number of video sensors, it is still difficult to accurately capture abnormal information in traffic video images in complex scenes (such as severe weather). It is one of the current application problems. As we all know, due to many complex factors such as vehicle exhaust and kerosene burning, smog has frequently occurred in most parts of my country in recent years, especially in the densely populated developed cities in the south. There are many particles and l...

AI Technical Summary

Problems solved by technology

However, on the one hand, the soft matting algorithm has a high computational load, and it is difficult to use it in real-time monitoring occasions; on the other hand, when the brightness of the scene object is similar to the atmospheric light, the restoration result of this method will appear obvious color Distortion (such as a large sky area), the loss of detailed information is relatively serious, and this situation is more common in traffic scenes, such as zebra crossings, etc.

Images